What does HackerNews think of netty-incubator-transport-io_uring?

Language: Java

Use io_uring for network I/O | Apr 2023
io_uring will indeed be a perfect fit for Java: no need to pay a supplemental JNI access cost for each IO as all you need is a memory barrier to read or write the shared queues which can be properly implemented in pure Java. We're not there yet but here is where you can look:

- For network I/O, Netty has an incubating transport that is promising [1].

- For disk I/O, JDK's Loom project [2] has mentioned its plan to rely on io_uring on Linux [3], but there's no ETA AFAIK.

[1] https://github.com/netty/netty-incubator-transport-io_uring

[2] https://openjdk.org/projects/loom/

[3] https://cr.openjdk.org/~rpressler/loom/loom/sol1_part1.html

Zero-copy network transmission with io_uring (2021) | Jan 2022
Expand Context ↕
Yes, I think io_uring is slowly making its way onto Java ecosystem. Example: https://github.com/netty/netty-incubator-transport-io_uring

I guess it will go into the JVM too.
Java Virtual Threads Preview | Nov 2021
Expand Context ↕
Maybe maybe not. If you want to take advantage of things like io_uring you're going to be doing that with a JNI lib. Such as this Netty incubator support for it https://github.com/netty/netty-incubator-transport-io_uring

Also all of the existing JVM IO is done with JNI. How do you think java.io is implemented itself? Of course Loom can change those implementations, but strictly speaking JNI is currently extremely common for Java IO. How big of a task supporting virtual threads for the Java libraries remains to be seen, especially for the unofficial extensions like the sun.nio.* package